Screening of newborns with growth curves derived from populations dissimilar to their own can lead to sizable errors. I present a technique for constructing percentile curves for birth weight at different gestational ages, using North Carolina's computerized birth certificate file. This method eliminates from the analysis most erroneous data on gestational age. The resulting intrauterine growth curves show weights at term for North Carolina newborns. These curves differ from previously reported norms in that weights are higher than those reported for two low-income hospital subpopulations, and lower than birth weights in populations with better living standards and prenatal care. Plotting North Carolina births on the Colorado intrauterine growth curve resulted in considerable overdiagnosis of "large for gestational age," while most infants who were small for gestational age failed to be identified.